[PATCH v2] s390/traps: Add exception statistics

[PATCH v2] s390/traps: Add exception statistics

[Heiko Carstens]

From: Sven Schnelle <svens@linux.ibm.com>

Add a new debugfs file which displays the number of exceptions (program
checks) per CPU. This is helpful for debugging purposes.

The statistics are typically available at
/sys/kernel/debug/s390/exceptions.

[ hca@linux.ibm.com: Forward ported code, changed file location ]

Suggested-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
---

Notes:
    v2:
    - Fix / remove broken iterator handling as reported by Sashiko:
      https://sashiko.dev/#/patchset/20260617113726.2079324-1-hca@linux.ibm.com?part=1

 arch/s390/kernel/traps.c | 41 +++++++++++++++++++++++++++++++++++++++-
 1 file changed, 40 insertions(+), 1 deletion(-)

diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c
index 564403496a7c..fcd8c1122ef5 100644
--- a/arch/s390/kernel/traps.c
+++ b/arch/s390/kernel/traps.c
@@ -9,7 +9,9 @@
  *    Copyright (C) 1991, 1992 Linus Torvalds
  */
 
+#include <linux/capability.h>
 #include <linux/cpufeature.h>
+#include <linux/debugfs.h>
 #include <linux/kprobes.h>
 #include <linux/kdebug.h>
 #include <linux/randomize_kstack.h>
@@ -33,6 +35,12 @@
 #include <asm/fault.h>
 #include "entry.h"
 
+struct pgm_stat {
+	unsigned int count[128];
+};
+
+static DEFINE_PER_CPU_SHARED_ALIGNED(struct pgm_stat, pgm_stat);
+
 static inline void __user *get_trap_ip(struct pt_regs *regs)
 {
 	unsigned long address;
@@ -332,6 +340,7 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
 	struct lowcore *lc = get_lowcore();
 	bool percpu_needs_fixup;
 	irqentry_state_t state;
+	struct pgm_stat *stat;
 	unsigned int trapnr;
 	union teid teid;
 
@@ -339,6 +348,10 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
 	regs->int_code = lc->pgm_int_code;
 	regs->int_parm_long = teid.val;
 	regs->monitor_code = lc->monitor_code;
+
+	trapnr = regs->int_code & PGM_INT_CODE_MASK;
+	stat = this_cpu_ptr(&pgm_stat);
+	stat->count[trapnr]++;
 	/*
 	 * In case of a guest fault, short-circuit the fault handler and return.
 	 * This way the sie64a() function will return 0; fault address and
@@ -383,7 +396,6 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
 	if (!irqs_disabled_flags(regs->psw.mask))
 		trace_hardirqs_on();
 	__arch_local_irq_ssm(regs->psw.mask & ~PSW_MASK_PER);
-	trapnr = regs->int_code & PGM_INT_CODE_MASK;
 	if (trapnr)
 		pgm_check_table[trapnr](regs);
 out:
@@ -393,6 +405,33 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
 	percpu_exit(regs, percpu_needs_fixup);
 }
 
+static int pgm_check_stat_show(struct seq_file *p, void *v)
+{
+	int i, cpu;
+
+	cpus_read_lock();
+	seq_puts(p, "          ");
+	for_each_online_cpu(cpu)
+		seq_printf(p, "CPU%-8d", cpu);
+	seq_putc(p, '\n');
+	for (i = 0; i < 128; i++) {
+		seq_printf(p, "%02x: ", i);
+		for_each_online_cpu(cpu)
+			seq_printf(p, "%10u ", per_cpu(pgm_stat, cpu).count[i]);
+		seq_putc(p, '\n');
+	}
+	cpus_read_unlock();
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(pgm_check_stat);
+
+static int __init debugfs_pgm_check_init(void)
+{
+	debugfs_create_file("exceptions", 0400, arch_debugfs_dir, NULL, &pgm_check_stat_fops);
+	return 0;
+}
+late_initcall(debugfs_pgm_check_init);
+
 /*
  * The program check table contains exactly 128 (0x00-0x7f) entries. Each
  * line defines the function to be called corresponding to the program check
-- 
2.53.0

Re: [PATCH v2] s390/traps: Add exception statistics

[Christian Borntraeger]

Am 17.06.26 um 14:33 schrieb Heiko Carstens:
> From: Sven Schnelle <svens@linux.ibm.com>
> 
> Add a new debugfs file which displays the number of exceptions (program
> checks) per CPU. This is helpful for debugging purposes.
> 
> The statistics are typically available at
> /sys/kernel/debug/s390/exceptions.
> 
> [ hca@linux.ibm.com: Forward ported code, changed file location ]
> 
> Suggested-by: Christian Borntraeger <borntraeger@linux.ibm.com>
> Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
> Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

Looks good to me
Tested-by: Christian Borntraeger <borntraeger@linux.ibm.com>


> ---
> 
> Notes:
>      v2:
>      - Fix / remove broken iterator handling as reported by Sashiko:
>        https://sashiko.dev/#/patchset/20260617113726.2079324-1-hca@linux.ibm.com?part=1
> 
>   arch/s390/kernel/traps.c | 41 +++++++++++++++++++++++++++++++++++++++-
>   1 file changed, 40 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c
> index 564403496a7c..fcd8c1122ef5 100644
> --- a/arch/s390/kernel/traps.c
> +++ b/arch/s390/kernel/traps.c
> @@ -9,7 +9,9 @@
>    *    Copyright (C) 1991, 1992 Linus Torvalds
>    */
>   
> +#include <linux/capability.h>
>   #include <linux/cpufeature.h>
> +#include <linux/debugfs.h>
>   #include <linux/kprobes.h>
>   #include <linux/kdebug.h>
>   #include <linux/randomize_kstack.h>
> @@ -33,6 +35,12 @@
>   #include <asm/fault.h>
>   #include "entry.h"
>   
> +struct pgm_stat {
> +	unsigned int count[128];
> +};
> +
> +static DEFINE_PER_CPU_SHARED_ALIGNED(struct pgm_stat, pgm_stat);
> +
>   static inline void __user *get_trap_ip(struct pt_regs *regs)
>   {
>   	unsigned long address;
> @@ -332,6 +340,7 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
>   	struct lowcore *lc = get_lowcore();
>   	bool percpu_needs_fixup;
>   	irqentry_state_t state;
> +	struct pgm_stat *stat;
>   	unsigned int trapnr;
>   	union teid teid;
>   
> @@ -339,6 +348,10 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
>   	regs->int_code = lc->pgm_int_code;
>   	regs->int_parm_long = teid.val;
>   	regs->monitor_code = lc->monitor_code;
> +
> +	trapnr = regs->int_code & PGM_INT_CODE_MASK;
> +	stat = this_cpu_ptr(&pgm_stat);
> +	stat->count[trapnr]++;
>   	/*
>   	 * In case of a guest fault, short-circuit the fault handler and return.
>   	 * This way the sie64a() function will return 0; fault address and
> @@ -383,7 +396,6 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
>   	if (!irqs_disabled_flags(regs->psw.mask))
>   		trace_hardirqs_on();
>   	__arch_local_irq_ssm(regs->psw.mask & ~PSW_MASK_PER);
> -	trapnr = regs->int_code & PGM_INT_CODE_MASK;
>   	if (trapnr)
>   		pgm_check_table[trapnr](regs);
>   out:
> @@ -393,6 +405,33 @@ void noinstr __do_pgm_check(struct pt_regs *regs)
>   	percpu_exit(regs, percpu_needs_fixup);
>   }
>   
> +static int pgm_check_stat_show(struct seq_file *p, void *v)
> +{
> +	int i, cpu;
> +
> +	cpus_read_lock();
> +	seq_puts(p, "          ");
> +	for_each_online_cpu(cpu)
> +		seq_printf(p, "CPU%-8d", cpu);
> +	seq_putc(p, '\n');
> +	for (i = 0; i < 128; i++) {
> +		seq_printf(p, "%02x: ", i);
> +		for_each_online_cpu(cpu)
> +			seq_printf(p, "%10u ", per_cpu(pgm_stat, cpu).count[i]);
> +		seq_putc(p, '\n');
> +	}
> +	cpus_read_unlock();
> +	return 0;
> +}
> +DEFINE_SHOW_ATTRIBUTE(pgm_check_stat);
> +
> +static int __init debugfs_pgm_check_init(void)
> +{
> +	debugfs_create_file("exceptions", 0400, arch_debugfs_dir, NULL, &pgm_check_stat_fops);
> +	return 0;
> +}
> +late_initcall(debugfs_pgm_check_init);
> +
>   /*
>    * The program check table contains exactly 128 (0x00-0x7f) entries. Each
>    * line defines the function to be called corresponding to the program check